The effect of population estimates on country-level nutrition data: Demographical gymnastics, nutritional conundrums

“There’s a fine line between the numerator and denominator. Only a fraction of you will get this," goes the old joke.

Working with data on the nutritional status of populations is a constant reminder of just how important denominators are – the prevalence of malnutrition and the success of any efforts to tackle it can only be measured if we know how many people comprise each of the groups we are interested in. Counting people is a tricky job, and the answer you get will depend on the methods and sources you use.

The most common source of country-level population data used at the global level is the UN Population Division’s World Population Prospects (WPP) compendium, which provides historic and current population estimates and future projections for the world, regions and countries. Most national censuses contribute to the WPP, which uses statistical modelling tools to calculate population figures based on censuses and other survey and surveillance data. The population estimates are used to calculate, among other indicators, the number and proportion of people affected by malnutrition in many countries at various time points. The WPP is updated biennially; estimates are often revised retrospectively in light of new data or more advanced statistical techniques.

What can happen to nutrition data when population estimates are revised retrospectively? Here’s an example:

Anaemia data for Kenya and Vanuatu

The problem

In the 2015 Global Nutrition Report (GNR) we reported that Burundi, Colombia, Kenya, Vanuatu, and Vietnam were on course to achieve reductions in anaemia among women of reproductive age – an assessment made using the WHO’s 2013 estimates of global and national trends in anaemia between 1995-2011, reported in the 2014 Joint Malnutrition Estimates (JME 2014). However, in the 2016 GNR, we reported that only Burundi, Columbia and Vietnam were on course, while Kenya and Vanuatu were no longer considered to be on course. Neither of these countries had published new surveys showing increases in anaemia—the reported prevalence of anaemia for Kenya and Vanuatu in the year 2011 was 25% and 21.7%, respectively, in both GNRs. Nor had the WHO published new modelled estimates showing changes in prevalence. So what happened?

The explanation

An assessment of “on course” or “off course” is based on the AARR – average annual rate of reduction – in anaemia. Countries with rates of reduction greater than 5.21% per year are deemed “on course”, meaning that they’ve shown a satisfactory rate of decline; an AARR of less than 5.21% means that they’re off course. In the 2015 GNR, the AARRs for Kenya and Vanuatu were 5.6% and 5.2%, respectively, and so they were among the five countries “on course” to meet the World Health Assembly (WHA) 2025 Maternal and Child Nutrition target of a 50% reduction in anaemia.

Later that year, the WPP released their 2015 revision with retrospectively revised population figures for all countries. The WHO group that work on the Joint Malnutrition Estimates (JME), of which anaemia data are a component, fine-tuned the population numbers for their estimates of anaemia for countries and published these in their JME 2015 update. The denominator “number of women of reproductive age” in Kenya and Vanuatu had therefore changed, while the estimated proportion of women with anaemia in the two countries had not. In turn, the numerator, “number of women of reproductive age who have anaemia”, also changed between the 2015 and 2016 GNRs—from 2.52 million to 2.49 million in Kenya, and from 13,300 to 13,000 in Vanuatu. A downward shift in the estimated number of women affected by anaemia pushed the AARRs lower during the re-adjustment, to 4.3% for Kenya and 4.4% for Vanuatu. These countries were now below the AARR threshold of 5.21% for “on course”, and were now re-classified as being “off course” for the same time period (2011) and same prevalence levels as a result of a retrospective change in underlying population sizes.

The consequences

What does this mean? For the two countries, it means a qualitative demotion in any assessment scorecards—like the GNR—that track their progress towards the WHA targets. Kenya and Vanuatu may not have reduced their efforts to tackle anaemia between 2015 and 2016, but were stripped of their erstwhile prize of being among the five countries doing well on this issue due to changes in statistical modelling techniques.

For those of us keeping an eye on global progress towards the WHA targets, it makes the picture a tad bleaker—only 3 of 193 countries in 2016 are making sufficient progress towards halving anaemia, a major public health problem that affects millions of women globally. But it also highlights the delicate balance of progress in global development: while these retrospective revisions to population estimates are often deemed an improvement in demographers’ ability to make strong predictions about population sizes, they can represent stagnation—or even regress—for the nutrition workforce trying to make the world a better-nourished place.

Note: thanks to Carrie Hubbell Melgarejo,
Nutrition Advisor, SPRING, for requesting a fuller explanation to this change in ‘on/off course’ status for Kenya and Vanuatu between the 2015 and 2016 GNRs.